Endpoint detection and response (EDR) watches activity on endpoints for signs of malicious behavior. It helps security teams detect threats, investigate what happened, and respond. Unlike traditional antivirus, which often looks for known malware signatures, EDR can look for suspicious behavior associated with threats that are harder to recognize. section 5.3.2 section 12.5.2
EDR is one part of a broader defense:
- During an attempted exploit, it can help identify suspicious processes or code injections. Patching, secure configuration, and least privilege still help prevent or limit the attack. section 12.2.4
- After a break-in, behavior-based detection can flag misuse of built-in utilities, even when no malicious file is written to disk. section 12.2.5
- EDR mainly sees endpoint activity. Network traffic, cloud systems, email, and identity activity may need other sources of visibility; extended detection and response (XDR) addresses this gap. section 12.5.2
A familiar tool used for the wrong purpose
PowerShell and Windows Management Instrumentation (WMI) are legitimate Windows utilities, but attackers can misuse them to run commands or maintain access. Behavior-based EDR looks for suspicious activity involving such tools rather than relying only on finding a malicious file. section 12.2.5
An EDR alert is a reason to investigate, not proof of an incident. A legitimate action can be flagged as malicious—a false positive. A real threat can also go unflagged—a false negative. Both possibilities matter when deciding what an alert means. section 5.3.2
|
Reality (what
really happened)
|
|||
|
Incident
|
No incident
|
||
|
Tool Detection Result
|
Incident
|
True Positive (TP)
Events
evidencing a real incident are correctly detected
|
False Positive (FP)
Benign
activity as an incident, leading to unnecessary analysis
|
|
No incident
|
False Negative (FN)
The tool
fails to detect a real incident, allowing the threats or malfunctions to go
unnoticed
|
True Negative (TN)
The tool
correctly identifies routine activity as non-malicious, avoiding unnecessary
alerts and dispersion of focus
|
|
Use the detection-results table to compare what a tool reported with what actually happened. It shows why an alert needs analysis and why no alert does not guarantee safety. section 5.3.2

